00:10 wayland left, wayland joined
apogee_ntv Updates pushed to MCP::Server, LLM::Chat (mostly around better tool calling) and ComfyUI::API (more control over cancellations & the queue). 00:32
Pushed Selkie 0.9.1, bunch of refactors, couple of bugfixes, removed some dead code. 00:46
01:08 rmv left 01:58 hulk joined 02:00 kylese left 02:06 leppard left 02:15 hulk left, kylese joined 02:48 lizmat_ joined 02:49 vrurg joined
disbot4 <aruniecrisps> @wayland having a real bad headache I'll get back to you on this 02:50
02:51 vrurg_ left, lizmat left 03:05 rmv joined, rmv left, rmv joined 03:09 rmv left 03:10 wayland left 03:12 apogee_ntv left, apogee_ntv joined 03:13 wayland76 joined 03:21 lichtkind__ joined 03:24 lichtkind_ left 04:06 guifa__ left 04:07 guifa joined 04:21 belluzj joined
wayland76 aruniecrisps: Sure, no worries. Apparently putting your feet/hands in the hottest water you can stand helps with some kinds of migranes (not "I'm about to burn", but "This is definitely warming me up a lot"). No promises, but might be worth a go. HTH, 04:23
Not medical advice for legal reasons :)
04:25 belluzj left 05:06 lichtkind joined 05:08 lichtkind__ left
disbot4 <vendethiel> Finallyyyyyy 05:17
<vendethiel> wayland76: it’s a different language/syntax for the Racket family of languages. There’s a quick intro page if you want a quick tour of what it looks like 05:18
<vendethiel> Spoiler: way less ()s than Racket itself :) 05:19
wayland76 Oh, nice! It looks vaguely rust-like. I was thinking about doing one someday called "Grammar Cookbook" with syntax inspired by a) grammar textbooks (ie. verbs instead of functions, various kinds of nouns, etc), and cookbooks (ie. data = ingredients, and recipe process = function). 06:00
(ie. an alternate syntax for Raku).
It's a long way down my list though, so I don't expect to start for 20 or so years. 06:01
Maybe a retirement project.
Thanks vendethiel (and aruniecrisps)
disbot4 <librasteve> hi @aruniecrisps I asked chat jippity to compare, TLDR; Raku’s grammar system is much more ambitious. chatgpt.com/share/6a3a27c0-7ab8-83...a05218778b 06:30
<librasteve> wayland76: have you seen www.amazon.com/Parsing-Perl-Regexe...1484232275 06:42
06:52 Sgeo left
wayland76 librasteve_: Yes, I went out a year or two ago and bought every Raku/Perl 6 book I could lay my hands on, partly as a donation to the community, and partly so I'd have them. Not sure I've read that one yet. 07:06
The Grammar/Actions are the only part of Raku internals I've had much of a look at. 07:07
Personally I'd like to see the grammars a) be reversible, and b) support streaming, but I suspect that's a ways away yet. 07:08
librasteve_: It also depends what question you ask. You may be interested in chatgpt.com/share/6a3a311b-786c-83...0c90001b9c (which is "What are the things that Rhombus grammars can do that Raku grammars don't.") 07:09
disbot4 <librasteve> interesting … I guess that would underpin things like code editors 07:10
wayland76 Yes, and network protocols.
disbot4 <librasteve> touché - btw when I share LLM chats I usually at least sniff check them for hallucinations 07:13
wayland76 Summary of "What can Rhombus do that Raku can't", the answer is "come back and talk to us when you have macros" :) . Also, Slangs might be easier.
Oh, good idea. I don't know enough about Racket/Rhombus to do that I think. Maybe I should've just shared the question :) . 07:14
disbot4 <librasteve> yours says If your goal is: * Extend the programming language itself * Add new declarations * Add new binding forms * Create new operators * Create a custom language that feels built-in then Rhombus’s macro/parsing system can do things that Raku grammars currently cannot. 07:15
<librasteve> of which Raku CAN do at least 3 :-)
wayland76 librasteve_: I'm currently guiding cursor through adding Data Oriented Programming (Trees + Tables) to Raku, which I think will be the most ambitious Slang yet (I'd like to see it in a future version of Raku, but think we should also test it out first).
disbot4 <librasteve> .oO 07:16
wayland76 Yes, though the caveat in the last paragraph explains why that's more of a difference in terminology (or rather, a poor question on my part) than an hallucination. 07:17
disbot4 <librasteve> that is one awesome use case for LLM … do you find it makes a super ambitious project like that doable by a one man band?
<librasteve> sorry must &afk for a bit 07:21
07:29 lizmat_ left, lizmat joined 07:32 abraxxa joined
wayland76 librasteve_: The problem is that my ambitions grew with it :p . The part that I'm working on now is essentially a query engine that adds the relevant querying to the language. It has very simple parsers/renderers for CSV/JSON. I'm about halfway through reviewing the code. I don't think it's great at deciding what the classes/roles should be, but when we redesign, it gets that done quickly. 07:34
I'm not sure that grammars are going to feature largely in the Grammar Cookbook language, because I'm expecting it to be pretty much all words, and only about the same level of punctuation that English has. Maybe a bit more, but not stacks. Formatting will be significant in some cases. My theory here is that, unlike developers, AIs don't care how many keystrokes it is, so readability will be king. 07:36
07:42 wayland joined, wayland76 left 07:55 dakkar joined 08:28 rmv joined 08:33 rmv left
apogee_ntv wayland: Isn't that always the way? :P Selkie started off as a todo app before I extracted a TUI framework from it. 08:57
Well GTD app
09:03 abraxxa left 09:05 abraxxa1 joined 09:32 leppard joined
wayland Haha, yeah :). We need more INTJ module developers. Someone said about INTJ people that "they try to solve a problem in such a way that no-one will ever need to solve it again". 09:42
apogee_ntv: Are you on the #mugs channel on IRC? 09:44
apogee_ntv: Despite the fact that it says it's "Game Development", most of the development actually seems to be TUI development, eg. Terminal::Widgets and the like. 09:57
10:01 rmv joined, rmv left, rmv joined 10:05 rmv left 10:56 camelia left 10:57 nine left 11:00 nine joined 11:11 wayland left, wayland joined 11:18 camelia joined
xinming wayland: I'm INTP, most time, I started something I feel interesting, after I get the whole picture of the project, and I know I can have it done, I lose my interest. 11:19
wayland xinming: haha, relatable. Something that might be helpful: 1. Investigate spec-driven development AIs, 2. Ensure that the way you get a picture of the whole project is by writing a specification (personally I write to learn about things, so that works for me), 3. Feed the specification to an AI that does spec-driven development. 11:41
The only reason I'm doing this Data-Oriented Programming thing is that most of the projects I do, I think "Data-Oriented Programming would be really handy here", and there aren't currently any data-oriented programming languages that I know of. XSLT does trees, and SQL/xHarbour do tables, but no-one does both. 11:42
I mean, you can do both data structures in any language, but it's like doing OOP in C.
disbot4 <simon_sibl> could someone explain to me how different is Raku start{} compared to Golang's go func() in terms of being lightweight and how it works simply ? 12:00
wayland I asked ChatGPT (because I'm not familiar with go). ChatGPT said that the Go one is lighter weight, In Raku, you do `my $promise = start { f() }; do-something-else(); await $promise;` -- the "start" returns a promise of future output, and then `await $promise` waits for the promised output. 12:15
The Raku one is moderately similar to the Javascript one, if you're familiar with that.
Is it just me, or is docs.perl6.org down? Or is the perl6 name really deprecated? 12:27
simon_sibl: Probably more useful if I'd done `my $result = await $promise`. 12:29
ugexe A go thread has a default stack size of like 8kb. A raku thread is probably closer to 8mb than 8kb 12:33
wayland librasteve_: I've found something that may be a problem. The raku docs link to raku.org/community/ (note the slash) and that's 404. If I visit raku.org/community (without the slash) then it works. Can I recommend we accept both URLs?
librasteve_ sure … actually I thought I had already fixed this … please can you raise an issue at github.com/Raku/raku.org/issues 12:40
timo a thread in MoarVM has, when it's first created, a smaller nursery than the default/maximum. it also corresponds to an operating system thread, so it also has a C stack, though there won't be terribly much on it most of the time 12:44
but `start { }` creates a job on the $*SCHEDULER, which by default is a ThreadPoolScheduler
the ThreadPoolScheduler keeps a pool of MoarVM threads that act as workers, pulling work from queues, plus a supervisor thread that keeps an eye on statistics to decide when it's time to start a new thread 12:45
what you get from `start { ... }` and what you call a goroutine are what's often called "green threads", because they are a lot smaller than operating system threads 12:46
13:05 rmv joined, rmv left, rmv joined 13:08 belluzj joined
wayland librasteve_: Sure, done. github.com/Raku/raku.org/issues/323 13:08
13:09 rmv left 13:11 leppard left
apogee_ntv wayland: I'm ENTP so... close? :D 13:27
timo not to be rude, but do people generally appreciate "i asked chatgpt and it said ..." style answers on chats? i've seen it be heavily frowned upon in places like mastodon, which are different enough from IRC that the answer may change 13:32
[Coke] It seems relevant to a technical discussion, giving the context of where the info came from. 13:34
tadzik I personally stop reading whenever I see "I asked an LLM and...", so it's a useful prefix 13:35
apogee_ntv It's useful to know because the information isn't as authoritative as i.e. a link to docs 13:37
LLMs are a useful place to get information but they are occasionally wrong.
tadzik that feels like a textbook definition of a bad source of information :) 13:39
timo oh no i don't mean "putting in front that you asked an llm", i mean posting answers from an LLM directly
you lot probably already know i'm vehemently LLM-opposed
apogee_ntv I'm largely the opposite, they're useful to me. Selkie would probably have taken twice as long to get anywhere without LLMs. 13:40
tadzik it feels a bit like "I asked this search engine and this is what the first result said; the source is known to be wrong, I don't know myself, and don't bother searching yourself because you likely won't get the same answer anyway". So I'm not sure what it adds, really
timo especially for topics that don't have bajillions of articles written about them already, like raku, LLMs tend to have not very much to draw from when generating answers
apogee_ntv Even when you account for the fact that I keep having to refactor everything. 13:41
timo my claim isn't that you can't be productive with that stuff 13:43
anyway, it's also probably off topic 13:44
but i'd still be interested to hear from more people in the chat(s) if a "i just took your question to an LLM and here's what it generated" is considered helpful and/or welcome
apogee_ntv For me sure 13:45
I mean I would probably have asked the LLM myself if I didn't know the answer
But if they can coax it to be more useful than I can, I wont complain :D
timo generally, it feels like the raku IRC channel + bridged discord is a place where you can expect an answer from someone who definitely knows in a relatively short timeframe, though timezone differences of course give you quite a spread of response times 13:47
apogee_ntv I have often asked an LLM for the different ways to do {paste C code} in other language. It's pretty good at that, even taking into account things like idioms and paradigms. 13:49
timo asking that kind of question in here, "how can I do X in raku like I did here in C" or whatever, is a great crystallization kernel for multiple people to converse and learn together 13:51
apogee_ntv I always worry about becoming a help vampire :P
I have experienced being on the other side of that
disbot4 <simon_sibl> ugexe: "A go thread has a default stack size of like 8kb. A raku thread is probably closer to 8mb than 8kb", thats a HUGE difference xD 13:52
timo simon_sibl, please read on for my explanations. that is not an apples-to-apples comparison at all 13:53
disbot4 <simon_sibl> I was wondering because I made a simple bittorrent client in Go, and I heavily used the goroutines and channels to exchange all the pieces between the workers but in Raku thats maybe a bad idea then
apogee_ntv I never really wrote much Go, did it for a few days and got annoyed at the verbosity 13:54
Makes C++ look terse and expressive.
disbot4 <simon_sibl> to me, its very readable and quite explicit, makes it easy to work in a team or with newbies, but I also enjoy Raku syntax of course (when its only me) 13:57
<simon_sibl> timo: I try to understand your explanation, but I am not amazingly knowledgable about all those, from what I understand from you is that Raku also has kind of green threads but for some reason its different than Golang because its linked to MoarVM threads which are actual OS threads, I thought Go also does this, "real" OS threads and it spreads the goroutines on them 13:58
timo yes, that's how green threads usually work. it's also called M-N Threading or M-to-N threading, because there are different numbers of tasks vs operating-system level threads 13:59
you can't run multiple different pieces of code at the same time without threads at the operating system level (as always, exceptions do exist)
I only mention MoarVM because MoarVM is what makes all of Rakudo go 14:00
disbot4 <simon_sibl> so since they both use green threads (M-N), what could make Raku's green threads more "heavy" than Raku goroutines ?
<simon_sibl> Golang goroutines* 14:01
timo rakudo is a heavier runtime in general than what go has. go can do a lot more at compile time, ahead of time. rakudo with moarvm has an optimizer that runs along-side the program and does its thing then
disbot4 <simon_sibl> I see, thank you ! 14:07
<simon_sibl> Currently its difficult for me to find a project to make in Raku, I really want to make something, but I dont know what. I feel that if its something I wanna share, I am better of going with Zig/Go which I can compile to static binary and have a rich stdlib. So currently I only make tiny scripts just for myself, which is nice but I would like to make something worth being called a "project". 14:09
<simon_sibl> latest thing I made that was "useful" for work was a client script that constantly send via udp the public IP (got from calling ifconfig.me) and a server script that gets the public IP and if it changed, would send an api request to the dns provider to update the record to point to that new IP, was nice and easy to make, too fast, the -Ofun lasted only during those 50 lines xD 14:14
14:21 abraxxa1 left 14:41 dakkar left, dakkar joined
timo it's always tricky to come up with project ideas if you don't have an obvious immediate itch to scratch, yeah 14:51
I have the "next best thing" to the public ip thing you have, where I just have a while True loop in fish that sends the output of `ip a l` to ssh to the server that also runs my irc client and cats it into a file 14:53
it helps to have a publicly routable ipv6 on your network interface, of course
disbot4 <melezhik.> Simon - are you looking for a new project involvement ? You may take a look at that - rockylinux.org/news/2026-06-08-get...-needs-you , Sparky part
<librasteve> omg I now have to worry I am becoming a help vampire (presumably that's an older person who is constantly trying to impress newbies and convery them) 15:07
timo I feel like it's good for everyone if it's commonplace and ordinary for people to say when they feel like checking out of a conversation for any reason, and it's understood that nothing unspoken / subliminal is going on when that happens, re the "help vampire" thing 15:10
disbot4 <librasteve> as to sharing LLM answers - I guess I have let that become a habit - sorry ... (in this case, I wanted to learn for myself how to explain the benefits of Raku Grammars vs. Racket Rhombus lexers since we are not going to convince many folks to try out Raku if we can't explain why its better than some other, more trendy, thing) 15:11
<librasteve> [and I thought by sharing that we could all become better proponents of Raku - well until Wayland hosed me down with some reality] 15:12
timo I haven't noticed anyone in particular doing the LLM answer thing, but I also don't read in here all too much any more
so i'm not actually sure what you're refering to here 15:13
disbot4 <librasteve> oh - I shared some LLM stuff here this morning 15:14
timo oh, you shared a link to chatgpt.com, is that it?
with a TLDR for the contents 15:15
15:15 xybre left
timo that has a distintcy different flavour to me than posting "i asked chatgpt your question and it said: bla bla bla bla" 15:15
disbot4 <librasteve> yep - well I can see that it could become annoying if done too much 15:16
timo it's a lot less worse than the other thing, in my opinion 15:17
disbot4 <librasteve> btw I patched Cro::WebApp
[Coke] wow is codeberg slow 15:21
timo it tends to get DDoS'd recently I think?
and it does occasionally actually go down 15:22
[Coke] ... /me wonders why iTerm on my mac had access to add calendar events. 15:30
15:32 sjn left, sjn joined 15:47 rmv joined, rmv left, rmv joined 16:27 belluzj left
disbot4 <comborico> Is there a live stream up for the conference? 16:28
16:30 cm_ joined 16:31 cm left, cm_ is now known as cm 16:34 dakkar left 16:35 kelt0m is now known as fdp 16:36 fdp is now known as Kelt0m
andinus timo: i think it's not useful because they can themselves ask the llm if they want to, i think it's noise. the point of asking here is to get something better than llm i would guess 17:03
17:11 rmv left 17:19 acidsys left 17:27 johnjay left 17:29 johnjay joined 17:34 human-blip left 17:35 johnjay left, human-blip joined 17:36 johnjay joined 17:42 acidsys joined, johnjay left 17:45 johnjay joined 17:51 johnjay left 17:58 johnjay joined 18:04 johnjay left 18:16 johnjay joined 18:37 rmv joined 18:41 rmv left 18:50 rmv joined, rmv left, rmv joined 19:13 itaipu left 19:28 itaipu joined 19:29 itaipu left 19:49 itaipu joined 20:28 dmvrtx left 20:32 dmvrtx joined 20:57 leppard joined 21:02 leppard left 22:12 Sgeo joined
wayland apogee_ntv: Unfortunately, looks like I'm the designated INTJ, but also unfortunately, I'm creating Data-Oriented Programming (Trees+Tables) first, so other things will have to wait :p ... 22:29
timo: For me, as someone who's done both, it very much depends. I chose to post one a link when I was discussing with librasteve above, but intentionally didn't post one for the question that simon_sibl asked. Sometimes ChatGPT surfaces interesting info that hasn't been written up anywhere. For example, I was asking about dynamic module loading, and it pointed me at CompUnit::RepositoryRegistry, which doesn't seem to be documented yet; when I asked about 22:37
doco, it said "There's nothing yet; I just read the source".
timo: The timezone thing may also be a factor in my answers. I've often found in the past that there's no-one awake when I am (I'm in Australia), and IRC can feel like shouting into the void at those times. I probably need to mentally adjust for the time of day, because some good answers were provided not long after I made mine (I'm not including mine in the good answers -- it was more "least bad answer to date"). 22:40
timo i feel like you'll find RepositoryRegistry itself useful very very very rarely 22:42
but not having anywhere to look that lets you know that is also bad
wayland timo: It probably would've simplified github.com/wayland/raku-Implementation-Loader if I'd known about RepositoryRegistry. 22:44
simon_sibl: Regarding projects, I usually ask "Am I processing text?" If so, it could well be the right tool for the job. 22:46
timo i'd have expected the sibling classes to be more helpful than the registry class
rakudo's compunit repo stuff is pretty powerful, but that makes it a bit tricky to figure out how the pieces fit together, and there are many 22:47
wayland librasteve_: No, that's the inverse of a help vampire -- you're the willing victim of help vampires :) . en.wiktionary.org/wiki/help_vampire
22:47 lichtkind left
timo librasteve_: I forgot to say I appreciate you patched Cro::WebApp, that's good 22:49
22:50 lichtkind joined
wayland librasteve_: Regarding me hosing you down with reality, I was more interested in "What can we do to make sure our grammars are the absolute best on the planet" :) . But good point. 22:50
timo I wish we had already reached the point where you can run grammars sensibly against unnormalized unicode, and easier access to switching between byte and codepoint view of characters 22:51
when you have to be compatible with network protocols or file formats that are notionally "unicode based" but combining characters aren't quite so magnetic as they are in raku with NFG 22:52
like when you put a combining diacritic of some kind at the start of a string in a json document
m: say qq["\c[COMBINING ACUTE ACCENT]hi"] # parsing this correctly as json in JSON::Tiny had to be achieved with ignoremark and then manually tearing off the combiners from the opening double quote 22:57
camelia "́hi"
wayland Sounds like there's a definite wishlist for Grammars then. 23:02
timo mostly my very personal nice-to-have 23:09
23:10 lucerne90 left 23:11 lucerne90 joined
Voldenet it'd be very useful to have for binary grammars too 23:25
wayland Yeah, I was thinking of binary grammars with the wishlist commeit. 23:38
*comment
Voldenet if it'd be possible to generate grammars out of kaitai it'd probably enable a whole range of processing - currently if I want to process binary in raku, I simply use cpp binary outputting text tokens out of binary files which is horrible 23:53

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!